Simplex Method and Sensitivity Analysis2012

Linear Programming:
Simplex Method and

Sensitivity Analysis
LP: Simplex Method and Interpretations

Some Important Questions about LP Model and Its Solutions:
What is the impact if additional resources available?

What if model parameters (cost/profit) change to LP solutions?
How to interpret LP solutions in managerial sense?
Standardization of LP through Slack/Surplus/Artificial variables:
1. Slack variables in "Resources" type constraints ():

How many resource i left in the solution? [S 0]
i
(e.g. X + 2X 40 X + 2X + S = 40)
1
2
1
2
1
If S =0, "resource" i fully utilized (critical), and
i
If S >0, some resource i are slack, available for other usage.
i
2. Surplus variables in "Demand" type constraints (>):
How much exceeds in "demand" i from Min-Requirement [Si 0]
(e.g. X1 + 2X2 40, X1 + 2X2 - S2 = 40)
If Si=0, demand i just meet minimum requirement, and
If Si>0, demand i has exceeded the minimum requirement.

3. Artificial variables in all ("" and "=") constraints: [ai 0]
To guarantee all variables 0 in solution searching process.
(e.g. X1 + 2X2 40, X1 + 2X2 - S2 + a2 = 40, or
(e.g. X1 + 2X2 = 40, X1 + 2X2 + a2 = 40)
Simplex Solution Method

Simplex is an iterative search algorithm for large LP problems, starting from
the initial ("origin", all X = 0) and moving toward adjacent "corner" points
at the direction in which improvement on objective function value is
maximized.
Key Property underlying Simplex: If one "corner" point solution is better than
all adjacent "corner" point solutions, it is "optimal".
Basic steps of Simplex Method:
Iteration Process: (Tabular Form)
Initial Point: all X = 0 (non-basic), all Si = bi (basic)

One Xi (with max-contribution) enter - one Si leave out.
Repeat until the "optimal" is reached ("stop" rule).
Special Cases in LP:
Degeneracy /Unbounded Solution /Unconstrained Variables

No Feasible Solution /Multi-Optimal /Redundant Constraints
All these cases will be "automatically" detected in Simplex method.
LP Solution: Managerial
Interpretations
How to interpret LP solutions in managerial sense?
Shadow (Dual) Price: the marginal contribution (opportunity cost) of
an additional unit of resource i to the objective function value
under the given Upper and Lower limits of resource.
Is it worth to obtain additional resources from outside?
For the resource that has a slack > 0, its shadow price = 0.
For the resource that has a slack = 0, its shadow price > 0.
In managerial sense, when the shadow price of resource i is greater than

the cost to obtain, it is worth to obtain.
Shift slack resources (their shadow price = 0) to other activities.
Trade-off for re-allocation of resources among activities.
Exchange resources with smaller shadow prices for those with

shadow prices.
larger
Graphical Interpretation: the resource constraints with a positive

shadow price are Active (critical) constraints (the optimal solution
pass through), while those with a zero shadow price are inactive
constraints.
Schematic
Presentatio
n of the
Simplex
Method
LP Solution: Sensitivity
Analysis
Sensitivity Analysis: concerns the impacts on optimal solution
X* = {x*j, j=1,..n) if one or more given model parameters are changed due to
market/supply/capacity/technology/.. changes.
Objective Function Coefficients - Cj (j=1,...n) may be changed:
Graphically, Changes in Cj - will change the slope of objective

function - which may or may not affect the optimal solution.
Managerial question: beyond what range of change (cost/profit)

(Lj Cj Uj) - X* (not Z) will be changed?
A larger range indicates insensitive of Z to changes in Cj.
Reduced Cost (for variables with xj = 0 in the final solution):

The amount of "cost" (In a MIN problem) to be reduced in order to make
this variable back to solution with a positive value. (Or the amount of
"profit" to be increased in a MAX problem). This information is also
shown in the "Upper Limit" (Allowable Increase - Uj) or "Lower Limit"
(Allowable Decrease - Lj) column in the solution printout.
Allowable Changing Range of

Coefficients
Resource Coefficients (RHS) - bi (i=1,...m) may be changed:
Graphically, changes in b - will parallel shift the constraint line i
which may or may not change the optimal solution:
change X* and Z if the constraint is active (Si = 0).

no change in X* and Z if the constraint is inactive (S i > 0).
Managerial question: beyond what range (BLi bi BUi) so X* will

be changed? (considering possible "loosening"/increase or
"tightening"/decrease in resource bi)
Resource consumption coefficients - aij may be changed:
Graphically, changes in aij - will change the slope of constraint line (or
parallel shift the constraint line up or down). The change will affect the
optimal solution if the related constraint is an active constraint (S i = 0).
Managerial question: beyond what range (ALij aij BUij) so X* will be
changed? (e.g., technological improvement or labor productivity
increase)
All of the Upper and Lower limits for Cj, bi, and aij will be given in the
Simplex solution.
When Adding or Deleting A Constraint: Will X*

change?
Adding a new constraint - have two possibilities:
1. A redundant (inactive) one, no impact on X*,
should remove it out.
(Graphically, the new constraint line passing through
outside of current feasible solution area.)
2.
Reducing feasible solution area, may change

X*, or not.
Otherwise, the feasible solution area is reduced, but

the X* remains unchanged.
If new constraint line cuts out the optimal solution
point from the original feasible solution area, the new
X* needs to be calculated.
(check by putting X* into new constraint)
Deleting an "old" constraint - also have two

possibilities:
If its slack Si = 0, active one, need to resolve problem.
If its slack Si > 0, inactive one, no impact on X*.
LP: Dual Formation

Duality: Prime vs. Dual: Counter-Expression of original LP
Duality: Symmetric and Reflexive relationship (a reflection of an object in
the "Mirror").
Max-Profit: by select Xi under resource constraints (bi),

Min-(opportunity) Cost - by select Yi under its contribution (cj).
Key Properties between Prime and its Dual: (Dual Theorem)
Same prime problem may have multiple dual formations.

Same problem, different Expression (Max Min).
Same optimal solution and objective function value.
The shadow prices of the prime is the decision variable value of the dual, and
vice verse.
(Principle of Complementary Slackness)
Formulation of the Dual from the Prime:
Standardize constraint set:
Max problem - all to () type /Min problem, all to ().

Replace equality constraint with two inequality ones.
Transforming standardized Prime to Dual following rules.
No. of Variables in Dual = No. of Constraints in Prime

No. of Constraints in Dual = No. of Variables in Prime
Cj in Prime bi in Dual / bi in Prime Cj in Dual
aij in Prime aji in Dual
Changes in the Technological

Coefficient for the High Note
Sound Company
The Primal-Dual Relationship
Write the dual of the following problem:

Min.
z = 5x1 + 2x1 3x3
Subject to;2x1 x2 + 3x2 60
x1 + 5x2 2x3 50
2x1 + x3
= 15
Standardize the problem as a minimization and

then write its dual.
Standardize:
Max z = -5x1 - 2x1 + 3x3
Subject to:
-2x1 + x2 - 3x2 -60
x1 + 5x2 - 2x3 50
2x2 + x3
15
-2x2 -
-15
x3
Dual
Min z= -60u1 + 50u2 + 15u3 15 u4
Subject to:
-2u1 + u2
-5
u1 + 5u2 + 2u3 2u4 -2

-3u1 -2u2 + u3 u4
Or:
Min z =-5x1 + 2x1 - 3x3
Subject to:
2x1-
x2 + 3x2 -60
-x1 - 5x2 - 2x3 50

2x2 + x3 15
-2x2 - x3 -15
Dual
Max z= 60u1 - 50u2 + 15u3 - 15 u4
Subject to:
2u1 - 1u2 + 0 u3 + 0u4 5
-u1 - 5u2 + 2u3 2u4 2
3u1 + 2u2 + u3 u4 -3
Or:
-3u1 -2u2 u3 +u4 3
Dual Formation: Economic

Interpretations
Why develop the Dual for the Prime Problem?
a) Exhibit important economic/managerial interpretation:
Managerial information can be obtained directly from dual
variables as from Shadow Price of resources in the Prime.
Dual model solution provides "opportunity cost" (Fair Price)
for the constraints in the Prime. This information may offer
many managerial insights for the original problem.
b) Provide basis for more efficient solution process.
The computational time of a LP problem is positively related
to the number of constraints in the model.
For a LP problem with more constraints than variables, its
dual problem has less constraints and more variables. In
such a case, it is computational efficient to solve the dual
problem than to solve the prime problem, especially for
large sized problems.
Duality has been widely used in many advanced MS models.
LP: Real World Applications
LP - is a very powerful, advanced, and complex MS

technique used in all different fields and disciplines.
LP and its Simplex solution method provide basis for other
advanced mathematical programming techniques (e.g.,
Integer Programming, Non-linear Programming).
There are many LP computer software packages available in
the market with different functions and capabilities.
Major issues in real world LP applications:
Deviations from LP assumptions (e.g.,
Linearity/Nonnegative).
Difficult in collecting data and accuracy of data.
There may be many different formulations for the same
problem. Specific formulation must be selected based on
the purpose of the study and managerial considerations.

Simplex Method and Sensitivity Analysis2012

Uploaded by

Copyright:

Available Formats

Simplex Method and Sensitivity Analysis2012

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Simplex Method and Sensitivity Analysis2012

Uploaded by

Copyright:

Available Formats

Linear Programming:

Simplex Method and

LP: Simplex Method and Interpretations

What is the impact if additional resources available?

Standardization of LP through Slack/Surplus/Artificial variables:

1. Slack variables in "Resources" type constraints ():

If Si>0, demand i has exceeded the minimum requirement.

Simplex Solution Method

Initial Point: all X = 0 (non-basic), all Si = bi (basic)

Special Cases in LP:

Degeneracy /Unbounded Solution /Unconstrained Variables

All these cases will be "automatically" detected in Simplex method.

Is it worth to obtain additional resources from outside?

In managerial sense, when the shadow price of resource i is greater than

Exchange resources with smaller shadow prices for those with

Graphical Interpretation: the resource constraints with a positive

Objective Function Coefficients - Cj (j=1,...n) may be changed:

Graphically, Changes in Cj - will change the slope of objective

Managerial question: beyond what range of change (cost/profit)

Reduced Cost (for variables with xj = 0 in the final solution):

Allowable Changing Range of

change X* and Z if the constraint is active (Si = 0).

Managerial question: beyond what range (BLi bi BUi) so X* will

When Adding or Deleting A Constraint: Will X*

Reducing feasible solution area, may change

Otherwise, the feasible solution area is reduced, but

Deleting an "old" constraint - also have two

LP: Dual Formation

Max-Profit: by select Xi under resource constraints (bi),

Key Properties between Prime and its Dual: (Dual Theorem)

Same prime problem may have multiple dual formations.

Formulation of the Dual from the Prime:

Standardize constraint set:

Max problem - all to () type /Min problem, all to ().

Transforming standardized Prime to Dual following rules.

No. of Variables in Dual = No. of Constraints in Prime

Changes in the Technological

The Primal-Dual Relationship

Write the dual of the following problem:

Standardize the problem as a minimization and

u1 + 5u2 + 2u3 2u4 -2

-x1 - 5x2 - 2x3 50

Dual Formation: Economic

LP: Real World Applications

LP - is a very powerful, advanced, and complex MS

You might also like